Auto-correlation of wafer characterization data and generation of composite wafer metrics during semiconductor device fabrication

Abstract

A system includes a controller with processors configured to execute an auto-correlation module embodied in one or more sets of program instructions stored in memory. The auto-correlation module is configured to cause the processors to receive one or more patterned wafer geometry metrics, receive wafer characterization data from one or more characterization tools, determine a correlation between the one or more patterned wafer geometry metrics and the wafer characterization data, generate a ranking of the one or more patterned wafer geometry metrics based on the determined correlation, construct a composite metric model from a subset of the one or more patterned wafer geometry metrics based on the ranking of the one or more patterned wafer geometry metrics, generate one or more composite wafer metrics from the composite metric model, and generate a statistical process control output based on the one or more composite wafer metrics.

What is claimed: 
     
       1. A system, comprising:
 a controller, wherein the controller includes one or more processors and memory configured to store one or more sets of program instructions, wherein the one or more sets of program instructions are configured to cause the one or more processors to:
 receive a plurality of patterned wafer geometry metrics for a wafer, wherein the plurality of patterned wafer geometry metrics includes at least two different patterned wafer geometry metrics; 
 receive wafer characterization data for the wafer from one or more characterization tools; 
 determine a correlation between the plurality of patterned wafer geometry metrics and the wafer characterization data with respect to die failure of the wafer; 
 generate an automatic ranking of the plurality of patterned wafer geometry metrics based on the determined correlation via at least one machine learning algorithm; 
 construct a composite metric model for the wafer from a subset of the plurality of patterned wafer geometry metrics based on the automatic ranking of the plurality of patterned wafer geometry metrics, wherein the subset of the plurality of patterned wafer geometry metrics includes a first metric with a highest ranking and at least a second metric with a second-highest ranking; 
 generate one or more composite wafer metrics for the wafer from the composite metric model, wherein the one or more composite wafer metrics have a correlation to die failure of the wafer that is greater than the correlation to die failure of the wafer of the plurality of patterned wafer geometry metrics; 
 generate a statistical process control output based on the one or more composite wafer metrics; and 
 adjust one or more characteristics of one or more process tools based on the statistical process control output to improve one or more fabrication processes. 
 
 
     
     
       2. The system of  claim 1 , wherein the one or more sets of program instructions are further configured to cause the one or more processors to:
 generate one or more correctables based on the one or more composite wafer metrics. 
 
     
     
       3. The system of  claim 2 , wherein the one or more sets of program instructions are further configured to cause the one or more processors to:
 provide the one or more correctables to the one or more process tools to improve the one or more fabrication processes. 
 
     
     
       4. The system of  claim 3 , wherein the one or more correctables are provided to the one or more process tools via a feedback loop. 
     
     
       5. The system of  claim 3 , wherein the one or more correctables are provided to the one or more process tools via a feed forward loop. 
     
     
       6. The system of  claim 1 , wherein the plurality of patterned wafer geometry metrics includes one or more nanotopography (NT) metrics. 
     
     
       7. The system of  claim 1 , wherein the plurality of patterned wafer geometry metrics includes one or more local shape curvature (LSC) metrics. 
     
     
       8. The system of  claim 1 , wherein the wafer characterization data includes wafer inspection data. 
     
     
       9. The system of  claim 1 , wherein the wafer characterization data includes wafer metrology data. 
     
     
       10. The system of  claim 9 , wherein the wafer metrology data includes inline wafer metrology data. 
     
     
       11. The system of  claim 1 , wherein the wafer characterization data includes electrical probe data. 
     
     
       12. The system of  claim 1 , wherein the correlation between the plurality of patterned wafer geometry metrics and the wafer characterization data includes a correlation coefficient, wherein a saturation level of the correlation coefficient is dependent on a number of wafer geometry metrics, wherein a prediction accuracy of the composite wafer metric is dependent on the level of saturation of the correlation coefficient. 
     
     
       13. The system of  claim 12 , wherein the correlation coefficient is an R 2 —value. 
     
     
       14. The system of  claim 12 , wherein the one or more sets of program instructions are further configured to cause the one or more processors to:
 determine the correlation coefficient between the plurality of patterned wafer geometry metrics and the wafer characterization data via the at least one machine learning algorithm. 
 
     
     
       15. The system of  claim 14 , wherein the at least one machine learning algorithm includes a classification algorithm. 
     
     
       16. The system of  claim 15 , wherein the classification algorithm includes logistic regression. 
     
     
       17. The system of  claim 15 , wherein the classification algorithm includes a decision tree. 
     
     
       18. The system of  claim 1 , wherein the one or more sets of program instructions are further configured to cause the one or more processors to:
 display the statistical process control output on a user interface coupled to the controller. 
 
     
     
       19. A system, comprising:
 one or more characterization tools; and 
 a controller, wherein the controller includes one or more processors and memory configured to store one or more sets of program instructions wherein the one or more sets of program instructions are configured to cause the one or more processors to:
 receive a plurality of patterned wafer geometry metrics for a wafer, wherein the plurality of patterned wafer geometry metrics includes at least two different patterned wafer geometry metrics; 
 receive wafer characterization data for the wafer from one or more characterization tools; 
 determine a correlation between the plurality of patterned wafer geometry metrics and the wafer characterization data with respect to die failure of the wafer; 
 generate an automatic ranking of the plurality of patterned wafer geometry metrics based on the determined correlation via at least one machine learning algorithm; 
 construct a composite metric model for the wafer from a subset of the plurality of patterned wafer geometry metrics based on the automatic ranking of the plurality of patterned wafer geometry metrics, wherein the subset of the plurality of patterned wafer geometry metrics includes a first metric with a highest ranking and at least a second metric with a second-highest ranking; 
 generate one or more composite wafer metrics for the wafer from the composite metric model, wherein the one or more composite wafer metrics have a correlation to die failure of the wafer that is greater than the correlation to die failure of the wafer of the plurality of patterned wafer geometry metrics; 
 generate a statistical process control output based on the one or more composite wafer metrics; and 
 adjust one or more characteristics of one or more process tools based on the statistical process control output to improve one or more fabrication processes. 
 
 
     
     
       20. A method comprising:
 receiving a plurality of patterned wafer geometry metrics for a wafer, wherein the plurality of patterned wafer geometry metrics includes at least two different patterned wafer geometry metrics; 
 receiving wafer characterization data for the wafer from one or more characterization tools; 
 determining a correlation between the plurality of patterned wafer geometry metrics and the wafer characterization data with respect to die failure of the wafer; generating an automatic ranking of the plurality of patterned wafer geometry metrics based on the determined correlation via at least one machine learning algorithm; 
 constructing a composite metric model for the wafer from a subset of the plurality of patterned wafer geometry metrics based on the automatic ranking of the plurality of patterned wafer geometry metrics, wherein the subset of the plurality of patterned wafer geometry metrics includes a first metric with a highest ranking and at least a second metric with a second-highest ranking; 
 generating one or more composite wafer metrics for the wafer from the composite metric model, wherein the one or more composite wafer metrics have a correlation to die failure of the wafer that is greater than the correlation to die failure of the wafer of the plurality of patterned wafer geometry metrics; 
 generating a statistical process control output based on the one or more composite wafer metrics; and 
 adjusting one or more characteristics of one or more process tools based on the statistical process control output to improve one or more fabrication processes. 
 
     
     
       21. The method of  claim 20 , further comprising:
 generating one or more correctables based on the one or more composite wafer metrics. 
 
     
     
       22. The method of  claim 21 , further comprising:
 providing the one or more correctables to the one or more process tools to improve the one or more fabrication processes. 
 
     
     
       23. The method of  claim 22 , wherein the one or more correctables are provided to the one or more process tools via a feedback loop. 
     
     
       24. The method of  claim 22 , wherein the one or more correctables are provided to the one or more process tools via a feed forward loop. 
     
     
       25. The method of  claim 20 , wherein the plurality of patterned wafer geometry metrics includes one or more nanotopography (NT) metrics. 
     
     
       26. The method of  claim 20 , wherein the plurality of patterned wafer geometry metrics includes one or more local shape curvature (LSC) metrics. 
     
     
       27. The method of  claim 20 , wherein the wafer characterization data includes wafer inspection data. 
     
     
       28. The method of  claim 20 , wherein the wafer characterization data includes wafer metrology data. 
     
     
       29. The method of  claim 28 , wherein the wafer metrology data includes inline wafer metrology data. 
     
     
       30. The method of  claim 20 , wherein the wafer characterization data includes electrical probe data. 
     
     
       31. The method of  claim 20 , wherein the correlation between the plurality of patterned wafer geometry metrics and the wafer characterization data includes a correlation coefficient, wherein a saturation level of the correlation coefficient is dependent on a number of wafer geometry metrics, wherein a prediction accuracy of the composite wafer metric is dependent on the level of saturation of the correlation coefficient. 
     
     
       32. The method of  claim 31 , wherein the correlation coefficient is an R 2 —value. 
     
     
       33. The method of  claim 31 , further comprising:
 determining the correlation coefficient between the plurality of patterned wafer geometry metrics and the wafer characterization data via the at least one machine learning algorithm. 
 
     
     
       34. The method of  claim 33 , wherein the at least one machine learning algorithm includes a classification algorithm. 
     
     
       35. The method of  claim 34 , wherein the machine learning algorithm includes logistic regression. 
     
     
       36. The method of  claim 34 , wherein the machine learning algorithm includes a decision tree. 
     
     
       37. The method of  claim 20 , further comprising:
 displaying the statistical process control output on a user interface.